Visualizing Deep Learning Models for Urban Health Analysis

Abstract

As humanity has become increasingly urbanized, physical and mental health problems have increased significantly among urban populations with a combined cost of treating these diseases estimated to be in the trillions of dollars. In parallel to these developments, a growing body of research suggests that the design of the built environment has significant correlations with both physical and mental health outcomes. This research, however, has been limited in its ability to make use of large remote sensing datasets to identify specific design features at the neighborhood scale that correlate with health outcomes. The development of methods that can efficiently find such correlations from ubiquitous remote sensing datasets, such as satellite images, would therefore allow researchers a greater level of insight into how specific urban planning and design features might relate to health. This research contributes knowledge on a novel mixed method workflow to address this issue.